Code-anchored context is documentation that points at the exact files it governs. Instead of a wiki page titled "Billing" that lives in a different tool and slowly loses its connection to reality, a code-anchored note declares which paths it covers — a glob like `src/billing/**` — so the system can tell, mechanically, which knowledge applies to which code.
This anchoring unlocks two things a wiki cannot do. First, retrieval by path: before anyone edits a file, the relevant decisions, invariants and gotchas for that exact file can be pulled automatically, because the matching is by glob intersection, not by someone remembering the right wiki page exists. Second, drift detection: when the anchored code changes, the note can be flagged for re-review, because the system knows the note and the change overlap.
The trade-off to manage is anchor breadth. An anchor that matches five to forty files describes a real concept; an anchor that matches a thousand files (`src/**`) describes nothing in particular and rots into a catch-all. Good code-anchored context is narrow and multi-anchor — one note per concept, pointed at the handful of places that concept actually lives.